No-reference Remote Sensing Image Quality Assessment Method Based On Natural Scene Statistics

Efficient No-reference image quality assessment is of great significance for the design and application of remote-sensing system.In this paper,three no-reference image quality methods are proposed for single distortion images,single distortion remote-sensing images,and multi-distortion remote-sensing images.(1)A NRIQA method based On multi-order gradients is proposed.A 18-dimensional image feature vector is constructed from gradient magnitude features,relative gradient orientation features and relative gradient magnitude features over 2 scales and 3 orders.The feature matrix and distortion types of known distorted images are used to train an AdaBoost BP neural network to determine the image distortion type;the feature matrix and subjective scores of known distorted images are used to train an AdaBoost BP neural network to determine distortion degree.Experimental results show that the proposed method has high accuracy in distortion type classification,high consistency with subjective scores and high robustness for all types of distortion.The performance of the proposed method is not constricted to a particular database and the proposed method has high operation efficiency.(2)A NRIQA method for remote-sensing images based on gradient weighted natural scene statistics in spatial domain is proposed.A 36-dimensional image feature vector is constructed from local normalized luminance features and gradient weighted LBP features of local normalized luminance map over 3 scales.The feature matrix and distortion types of known distorted remote-sensing images are used to train a SVM classifier to determine the image distortion type;on the basis of SVM classifier,the feature matrix and subjective scores of known distorted remote-sensing images are used to train a SVM grader to determine distortion degree.Experimental results show that the proposed method has high accuracy in distortion type classification,high consistency with subjective scores and high robustness for all types of distortion of remote sensing images.The performance of the proposed method is not constricted to a particular database and the proposed method has high operation efficiency.(3)A NRIQA method for multiply distortion remote-sensing images based on deep convolutional neural network is proposed.First,the local contrast normalization and block process are carried out.Then,the information entropy threshold is used to remove the image patches with less information.Finally,a deep CNN model with two convolution layers is used to extract feature and determine distortion degree.Experimental results show that the proposed method has high consistency with subjective scores for multi-distortion remote-sensing images and common multi-distorted images.The performance of the proposed method is not constricted to a particular database and the proposed method has high operation efficiency.